An acquisition and a modulated recognition system for driver profiling in Malaysia / Ward Ahmed Alaulddin Al-Hussein
The process of collecting driving data and using a computational model to generate a safety score for the driver is known as driver profiling. Previous studies in Malaysia relied on simulators, questionnaires, and surveys to collect driving data. Such methods were criticized for being biased and unt...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2022
|
| Subjects: | |
| Online Access: | http://studentsrepo.um.edu.my/14509/1/Ward_Ahmed.pdf http://studentsrepo.um.edu.my/14509/2/Ward_Ahmed_Alaulddin.pdf http://studentsrepo.um.edu.my/14509/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The process of collecting driving data and using a computational model to generate a safety score for the driver is known as driver profiling. Previous studies in Malaysia relied on simulators, questionnaires, and surveys to collect driving data. Such methods were criticized for being biased and untrustworthy. Furthermore, due to the disparity in traffic laws and regulations between countries, what is deemed aggressive behavior in one place may not be the same in another. As a result, adopting existing profiles is not ideal. This research presents the first naturalistic driving study in Malaysia, in which thirty drivers were recruited to drive an instrumented vehicle for an accumulated distance of 750 kilometers. The data acquisition system consisted of various sensors, including On-Board Diagnostics II (OBDII), lidar, ultrasonic sensors, Inertial Measurement Unit (IMU), and Global Positioning System (GPS). The collected data were then utilized to establish credible driver profiles based on criteria developed in consultation with traffic experts. Following that, three deep-learning-based algorithms, namely, Deep Neural Network (DNN), Recurrent Neural Network (RNN), and Convolutional Neural Network (CNN), were modulated to classify the recorded driving data according to the established profiles. The results have shown that CNN outperformed the other two classification algorithms in terms of accuracy, precision, recall, and f-measure and was therefore selected for a recognition system that, in combination with the acquisition system, would assist traffic police and insurance firms in detecting unsafe driving behaviors. Furthermore, the study examined the effects of various factors on driving in Malaysia. The statistical results revealed that driving behavior is greatly influenced by drivers’ gender, age, and cultural background. There were also significant behavioral differences between those who drove on weekends and those who drove on weekdays. Finally, several recommendations were presented to government agencies based on the findings to improve road safety and help avoid future accidents.
|
|---|